The functionality was announced at the Interrupt conference. Before this update, locating these insights required developers to construct custom evaluators or manually scan through hours of conversation data to identify behavioral patterns in LLM applications.
